Podcast: AI Agents, MCP and Conversational Analytics
Johan Strand joins Tony Hammarlund to explore how AI agents, Conversational Analytics and MCP are changing the way marketers and analysts interact with their data.
What happens when working with analytics data starts to look less like building reports and writing queries, and more like having a conversation?
In this episode of Digital Marknadsföring med Tony Hammarlund, Ctrl Digital partner and senior digital analyst Johan Strand returns to the expert panel to discuss the rapidly developing landscape of AI-assisted analytics.
The conversation covers Google’s Ask Advisor, Conversational Analytics and Data Agents, as well as the growing role of Model Context Protocol (MCP) in connecting tools such as Claude and ChatGPT to external platforms and data.
Different Ways of Talking to Your Data
There is no single approach to AI-assisted analytics.
Platform-specific agents such as Google’s Ask Advisor can provide a convenient way to interact with data inside a particular ecosystem. Conversational Analytics and Data Agents provide another approach, allowing organisations to create agents with their own context and guardrails around data.
MCP opens up a different model by enabling general-purpose AI tools to connect to external systems and data sources.
Johan and Tony compare these approaches and discuss where each can be useful, as well as their limitations.
The Data Foundation Still Matters
Making it easier to ask questions does not automatically make the answers reliable.
A recurring theme throughout the episode is the importance of the underlying data setup and context. AI can make analytics more accessible, but it still needs to understand what the data represents, how metrics should be interpreted and which analytical rules should be applied.
This makes existing technical debt increasingly important. Inconsistent tracking, fragmented data, unclear definitions and missing business context become constraints when an AI system is expected to analyse data rather than simply display it.
The discussion also distinguishes between reporting and analysis. Retrieving a metric or identifying a change is a more constrained task than determining why something happened or recommending what should happen next. More complex analysis requires additional context, validation and, in many cases, a human in the loop.
At the same time, AI opens up useful new workflows. These range from quickly querying quantitative data to analysing qualitative information, verifying results and running scheduled processes such as anomaly detection.
In This Episode
Johan and Tony discuss:
- Google’s Ask Advisor and other platform-specific AI agents
- Conversational Analytics and Data Agents
- Adding context and guardrails to AI-assisted analytics
- How MCP connects AI tools with external platforms and data
- The difference between reporting and analysis
- Where AI is currently most useful in analytics workflows
- Why technical debt becomes a barrier to AI adoption
- The importance of human validation
- Scheduled analysis and anomaly detection
- Meridian Studio and marketing mix modeling
- Changes to Google Tag Manager
- Google Ads Data Manager and Microsoft Clarity
Listen to the Episode
The podcast Digital Marknadsföring med Tony Hammarlund features interviews and expert discussions with specialists from across the Swedish digital marketing industry.
You can find the episode everywhere podcasts are found. But here are links to the major platforms: